CN-121984868-A - Consumption finance mobile terminal rendering scheduling method based on bandwidth prediction

Abstract

The invention relates to the technical field of computer networks, and discloses a rendering scheduling method of a consumer financial mobile terminal based on bandwidth prediction. The method comprises the steps of continuously collecting downlink bandwidth at a mobile terminal side, obtaining local change rate based on difference and moving average, predicting a lower period target bandwidth by combining window average bandwidth, performing lower limit correction, dividing a terminal page into multiple areas, presetting minimum and maximum configuration of the total threads and each area, realizing fine distribution according to priority by mapping the target bandwidth and GPU threads, combining resource compression, expansion and integer correction, periodically collecting rendering start and stop moments of each area, constructing a response time sequence, cutting off abnormity, calculating time delay and performance alarm, adaptively adjusting bandwidth sampling period and sampling point number according to prediction deviation, forming a measurement-prediction-allocation-evaluation-optimization closed loop, and improving bandwidth prediction precision, resource utilization rate and rendering stability.

Inventors

• Lv Lingfeng
• XU HUI
• LU XUEZHENG

Assignees

• 浙江宁银消费金融股份有限公司

Dates

Publication Date

20260505

Application Date

20251225

Claims (9)

1. 1. The rendering scheduling method of the consumer-finance mobile terminal based on bandwidth prediction is characterized by comprising the following steps: Executing time-sharing acquisition and initialization construction of network bandwidth data at a mobile terminal side to form an available downlink bandwidth sampling sequence corresponding to each scheduling period; Based on the available downlink bandwidth sampling sequence, calculating and smoothing the bandwidth change rate of the adjacent sampling points to obtain a final estimated value of the local bandwidth change rate of the current scheduling period; calculating average available bandwidth according to bandwidth sampling values in a bandwidth acquisition time window, and executing prediction and lower limit correction on a lower period target bandwidth by combining a local bandwidth change rate final estimated value; Dividing a plurality of resource areas in a terminal display page, setting a total amount of resources, and setting a minimum configuration value and a maximum configuration value for each resource area to form multi-area resource allocation constraint and resource demand description; Constructing a mapping relation between the bandwidth and the rendering thread number of the graphic processor in each resource area according to the target bandwidth of the lower period, and executing segmentation processing on the continuous resource demand number of each resource area; under the constraint of the total amount of resources, executing resource compression, expansion and allocation correction on the continuous resource demand quantity of each resource region acquired according to the mapping relation to obtain the rendering thread integer allocation quantity of each resource region graphics processor meeting the constraint condition; Collecting rendering timing starting time and rendering timing ending time of each resource region, constructing a response time sequence, executing response time truncation and average response time delay calculation, and obtaining region rendering response characteristics and dynamic delay measurement results; And periodically executing bandwidth prediction deviation statistics and evaluation, and adjusting the bandwidth sampling period and the number of bandwidth sampling points in the subsequent scheduling period according to the evaluation result to finish scheduling period evaluation and sampling density reconfiguration state delay measurement results.
2. 2. The rendering scheduling method of a consumer-finance mobile terminal based on bandwidth prediction according to claim 1, wherein the performing time-sharing acquisition and initialization construction of network bandwidth data on the mobile terminal side forms an available downlink bandwidth sampling sequence corresponding to each scheduling period, and specifically comprises: The network monitoring module is arranged in the mobile terminal, selectable values of a preset bandwidth sampling period are 0.2 second and 0.1 second, and the bandwidth sampling period of 0.2 second is adopted in a first scheduling period when the system is started; Numbering the scheduling periods, recording the starting time of each scheduling period as the starting time of the corresponding scheduling period, setting a bandwidth acquisition time window with the duration of 10 seconds in each scheduling period, executing a plurality of continuous available downlink bandwidth sampling operations in the current bandwidth acquisition time window according to the preset bandwidth sampling period, acquiring all bandwidth sampling values in the current bandwidth acquisition time window, and taking the sampling sequence numbers as discrete sampling indexes; Setting the starting time of the next scheduling period as the starting time of the previous scheduling period plus the length of the bandwidth acquisition time window between the adjacent scheduling periods, so that the bandwidth acquisition time windows of the scheduling periods are continuously arranged on a time axis; in each scheduling period, according to the discrete sampling index, constructing all bandwidth sampling values acquired in time sequence into an available downlink bandwidth sampling sequence, wherein the available downlink bandwidth sampling sequence covers all sampling points in a bandwidth acquisition time window of the corresponding scheduling period; An integrity check is performed on each bandwidth sample value in the sequence of available downstream bandwidth samples, and when any sample value is detected to be less than zero, the sample value is corrected to zero.
3. 3. The rendering scheduling method of a consumer-finance mobile terminal based on bandwidth prediction according to claim 2, wherein the calculating and smoothing processes are performed on the bandwidth change rate of the adjacent sampling points based on the available downlink bandwidth sampling sequence to obtain a final estimated value of the local bandwidth change rate of the current scheduling period, and the method specifically comprises: for an available downlink bandwidth sampling sequence in each scheduling period, performing differential operation on bandwidth values of two adjacent samples in the sequence, and calculating the bandwidth change rate of each sampling point according to the proportional relation between the bandwidth difference value and the bandwidth sampling period to form a bandwidth change rate sequence covering all sampling points in the current scheduling period; In the bandwidth change rate sequence, constructing a moving average operation window by taking the latest 5 continuous change rate samples as a group, and executing arithmetic average operation on the bandwidth change rate samples in each window to obtain a local bandwidth change rate initial estimated value of the current scheduling period; Setting the absolute upper limit of the bandwidth change rate to 25000 kilobits per second, performing saturation truncation processing on the local bandwidth change rate initial estimated value, correcting the local bandwidth change rate initial estimated value to be a positive upper limit when the local bandwidth change rate initial estimated value is larger than the positive upper limit, correcting the local bandwidth change rate initial estimated value to be a negative upper limit when the local bandwidth change rate initial estimated value is smaller than the negative upper limit, and keeping the local bandwidth change rate initial estimated value unchanged under the rest conditions to obtain the local bandwidth change rate final estimated value of the current scheduling period.
4. 4. The rendering scheduling method of a consumer-finance mobile terminal based on bandwidth prediction according to claim 3, wherein the calculating average available bandwidth according to bandwidth sampling values in a bandwidth acquisition time window, and combining with a final estimated value of local bandwidth change rate, performs prediction and lower limit correction on a lower period target bandwidth, specifically comprises: In each scheduling period, taking a bandwidth acquisition time window corresponding to the scheduling period as a range, and executing arithmetic average operation on all bandwidth sampling values in the bandwidth acquisition time window to obtain the average available bandwidth of the current scheduling period; Calculating a predicted average available bandwidth of a next scheduling period according to a linear extrapolation relationship between the length of the scheduling period and the local bandwidth change rate based on the average available bandwidth of the current scheduling period and the local bandwidth change rate final estimated value; and when the predicted average available bandwidth of the next scheduling period is smaller than zero, correcting the predicted average available bandwidth to be zero as the target bandwidth of the next period.
5. 5. The rendering scheduling method of a consumer-finance mobile terminal based on bandwidth prediction according to claim 4, wherein the steps of dividing a plurality of resource areas in a terminal display page, setting a total amount of resources, setting a minimum configuration value and a maximum configuration value for each resource area, and forming a multi-area resource allocation constraint and a resource demand description include: Dividing the logic layout of a terminal display page into three resource areas, wherein the first resource area is a transaction function area, the second resource area is a data information display area, and the third resource area is an advertisement and recommendation module area; Setting a total amount of resources in the rendering threads of the graphics processor currently available to the terminal, and defining the total amount of resources as the total number of the rendering threads of the graphics processor, which can be distributed in a single scheduling period, of three resource areas; Setting a minimum configuration value and a maximum configuration value of the number of graphics processor rendering threads for each resource region respectively, wherein the minimum configuration value of the first resource region is 60 graphics processor rendering threads, the maximum configuration value of the first resource region is 80 graphics processor rendering threads, the minimum configuration value of the second resource region is 20 graphics processor rendering threads, the maximum configuration value of the second resource region is 30 graphics processor rendering threads, the minimum configuration value of the third resource region is 10 graphics processor rendering threads, the maximum configuration value of the third resource region is 20 graphics processor rendering threads, and the number of graphics processor rendering threads obtained by each resource region is limited between the corresponding minimum configuration value and the maximum configuration value in any scheduling period.
6. 6. The method for rendering and scheduling the consumer-finance mobile terminal based on bandwidth prediction according to claim 5, wherein the constructing the mapping relationship between the bandwidth and the number of rendering threads of the graphics processor of each resource region according to the lower period target bandwidth, and executing the segmentation processing on the number of continuous resource demands of each resource region, specifically comprises: Setting a reference bandwidth of a mapping relation between bandwidth and the number of rendering threads of a graphics processor to be 5000 kilobits per second, after predicting a target bandwidth of a next scheduling period, performing linear interpolation operation on a difference value between a minimum configuration value and a maximum configuration value of each resource region according to a ratio between the target bandwidth of the next period and the reference bandwidth, and calculating to obtain the number of continuous resource demands of each resource region of the next scheduling period, wherein the number of continuous resource demands is represented by the number of rendering threads of the graphics processor; When the continuous resource demand quantity of any resource region is smaller than the minimum configuration value of the corresponding resource region, the continuous resource demand quantity of the corresponding resource region is corrected to the minimum configuration value of the corresponding resource region, and the segmentation correction of the bandwidth and the resource mapping result is completed.
7. 7. The method for rendering and scheduling the consumer-finance mobile terminal based on bandwidth prediction according to claim 6, wherein the executing the resource compression, the expansion and the allocation correction on the continuous resource demand quantity of each resource area obtained according to the mapping relation under the constraint of the total amount of resources to obtain the integer allocation quantity of rendering threads of each resource area graphics processor meeting the constraint condition specifically comprises: Summing the continuous resource demand quantity of each resource region among the three resource regions for the next scheduling period to obtain the total resource application quantity of the next scheduling period; When the total resource application amount is less than or equal to the total resource amount, executing the following steps: S101, setting the continuous allocation quantity of rendering threads of a basic graphics processor of each resource area as the continuous resource demand quantity of the corresponding resource area; S102, subtracting the sum of the continuous allocation amounts of the rendering threads of the three resource area basic graphics processors from the total amount of resources, and calculating the number of the rendering threads of the graphics processors which can be allocated; S103, calculating the maximum complement quantity which can be added between the continuous allocation quantity of the basic graphic processor rendering threads and the maximum configuration value of the first resource area aiming at the first resource area, when the number of the residual allocatable graphic processor rendering threads is smaller than or equal to the maximum complement quantity, allocating all the residual graphic processor rendering threads to the first resource area, when the number of the residual allocatable graphic processor rendering threads is larger than the maximum complement quantity, setting the complement quantity of the first resource area as the maximum complement quantity, and updating the number of the residual allocatable graphic processor rendering threads to be the original residual allocatable graphic processor rendering threads minus the maximum complement quantity; S104, calculating the maximum complement quantity of the second resource area according to the continuous allocation quantity and the maximum configuration value of the rendering threads of the basic graphic processor of the second resource area, adding the complement quantity to the second resource area under the constraint of the updated number of the rendering threads of the residual graphic processor, and updating the number of the rendering threads of the residual graphic processor into the new number of the rendering threads of the residual graphic processor; S105, calculating the maximum complement quantity of the third resource region according to the continuous allocation quantity and the maximum configuration value of the basic graphics processor rendering threads of the third resource region, and adding the complement quantity to the third resource region under the constraint of the latest residual allocable graphics processor rendering threads, so that the sum of the continuous allocation quantity of the graphics processor rendering threads respectively obtained by the three resource regions after the complement is equal to the total resource quantity; when the total resource application amount is larger than the total amount of resources, executing the following steps: S201, adding the minimum configuration values of the three resource areas to obtain the sum of the lower limits of the three resource areas; S202, calculating the difference value between the continuous resource demand quantity and the minimum configuration value of the corresponding resource region aiming at each resource region to obtain the resource allowance of each resource region above the lower limit of the resource, and summing the resource allowance of the three resource regions to obtain the total resource allowance; S203, calculating a unified compression ratio coefficient according to the total amount of resources, the sum of lower limits of resources and the total resource allowance, multiplying the resource allowance of each resource area above a minimum configuration value by the unified compression ratio coefficient, and adding the minimum configuration value of the corresponding resource area to obtain a new continuous allocation amount of rendering threads of the graphics processor in each resource area; Performing downward rounding operation on the new graphics processor rendering thread continuous allocation amount of each resource area, obtaining an integer part as an initial graphics processor rendering thread integer allocation amount of the corresponding resource area, subtracting the decimal part corresponding to the corresponding resource area from the initial graphics processor rendering thread integer allocation amount of the new graphics processor rendering thread continuous allocation amount, summing the initial graphics processor rendering thread integer allocation amounts of the three resource areas, and subtracting the sum of the initial graphics processor rendering thread integer allocation amounts of the three resource areas from the total sum of the resources to obtain the number of residual allocable graphics processor rendering threads; When the number of the residual allocatable graphic processor rendering threads is zero, setting the integer allocation amount of the final graphic processor rendering threads of each resource area as the corresponding integer allocation amount of the initial graphic processor rendering threads; when the number of the residual allocatable graphic processor rendering threads is one, comparing according to the fractional part sizes of the three resource areas, selecting the resource area with the largest fractional part, increasing the integer allocation of the initial graphic processor rendering thread of the resource area with the largest fractional part by one graphic processor rendering thread, and keeping the integer allocation of the final graphic processor rendering thread of the residual resource area as the integer allocation of the initial graphic processor rendering thread of each; When the number of the rendering threads of the residual allocatable graphics processor is two, sorting according to the sizes of the decimal parts of the three resource areas, selecting the largest resource area of the decimal part as a first priority area, selecting the next largest resource area of the decimal part as a second priority area, adding one graphics processor rendering thread to the integer allocation amount of the initial graphics processor rendering threads of the first priority area and the second priority area respectively, and keeping the integer allocation amount of the final graphics processor rendering threads of the residual resource areas as the integer allocation amount of the corresponding initial graphics processor rendering threads.
8. 8. The bandwidth prediction-based rendering scheduling method of a consumer-finance mobile terminal according to claim 7, wherein the steps of collecting rendering timing starting time and rendering timing ending time of each resource region, constructing a response time sequence, executing response time truncation and average response time delay calculation, and obtaining region rendering response characteristics and dynamic delay measurement results include: Setting the number of samples of the rendering response time measurement to be 10 for each resource region in a single scheduling period; Recording a rendering timing starting time and a rendering timing ending time for each response time sample in each resource area, and calculating the response time of the test by subtracting the rendering timing starting time from the rendering timing ending time; setting the cut-off upper limit of the response time to 2000 milliseconds, correcting the response time sample to zero when the calculated response time is smaller than zero, correcting the response time sample to the cut-off upper limit when the calculated response time is larger than the cut-off upper limit, and keeping the calculated response time value unchanged under the other conditions; For each resource region, collecting all 10 response time samples in one scheduling period, constructing a response time sequence of the corresponding resource region in the corresponding scheduling period, and executing arithmetic average operation on all samples in the response time sequence to obtain average response time delay of the corresponding resource region in the corresponding scheduling period; Setting the response time performance judging threshold to 200 milliseconds, generating a performance warning event mark for the corresponding resource region in the corresponding scheduling period when the average response time delay of any resource region in a certain scheduling period is larger than the response time performance judging threshold, and not generating the performance warning event mark when the average response time delay is not larger than the response time performance judging threshold.
9. 9. The method for rendering and scheduling the consumer-finance mobile terminal based on bandwidth prediction according to claim 8, wherein the periodically executing bandwidth prediction bias statistics and evaluation, adjusting the bandwidth sampling period and the number of bandwidth sampling points in the subsequent scheduling period according to the evaluation result, and completing the scheduling period evaluation and the sampling density reconfiguration state delay measurement result, specifically comprises setting an evaluation time interval to 60 seconds, and executing a bandwidth prediction bias statistics and sampling period reconfiguration process when the time difference between the starting time of a certain scheduling period and the starting time of a first scheduling period is an integral multiple of the evaluation time interval; When bandwidth prediction deviation statistics is executed, determining the number of historical scheduling periods participating in the statistics according to the number of the currently completed evaluation time intervals, setting the number of the historical scheduling periods participating in the statistics as the number of the scheduling periods preceding the current scheduling period when the number of the scheduling periods preceding the current scheduling period is not more than 6, and setting the number of the historical scheduling periods participating in the statistics as 6 when the number of the scheduling periods preceding the current scheduling period is more than 6; When the number of the historical scheduling periods participating in statistics is larger than zero, the number of the historical scheduling periods participating in statistics is traced back from the current scheduling period, for each historical scheduling period, a relative prediction deviation sample of the historical scheduling period is calculated according to an average available bandwidth real measurement value of the historical scheduling period and a prediction average available bandwidth calculated for the historical scheduling period in the previous scheduling period, and a deviation sample sequence containing the relative prediction deviation samples of all the historical scheduling periods is constructed; when the number of historical scheduling periods participating in statistics is equal to zero, a deviation sample sequence is not constructed, and the statistical value of the average relative prediction deviation is set to zero; in the case of a deviation sample sequence, performing arithmetic average operation on all the relative prediction deviation samples in the deviation sample sequence to obtain average relative prediction deviation of the current scheduling period; Setting a prediction deviation judging threshold value to 0.3, when the number of historical scheduling periods participating in statistics is equal to zero, keeping a bandwidth sampling period to 0.2 seconds, when the number of historical scheduling periods participating in statistics is greater than zero and the average relative prediction deviation is greater than the prediction deviation judging threshold value, adjusting the bandwidth sampling period to 0.1 seconds, and when the number of historical scheduling periods participating in statistics is greater than zero and the average relative prediction deviation is not greater than the prediction deviation judging threshold value, setting the bandwidth sampling period to 0.2 seconds; After the bandwidth sampling period is updated, calculating the number of bandwidth sampling points in a subsequent scheduling period according to the ratio between the length of the bandwidth acquisition time window and the updated bandwidth sampling period, and sequentially executing the processing procedures of bandwidth acquisition, bandwidth change rate modeling, bandwidth prediction, resource mapping and allocation correction and region rendering response evaluation and sampling density reconfiguration in the next scheduling period according to the updated bandwidth sampling period and the number of bandwidth sampling points.

Description

Consumption finance mobile terminal rendering scheduling method based on bandwidth prediction Technical Field The invention relates to the technical field of computer networks, in particular to a rendering scheduling method of a consumer financial mobile terminal based on bandwidth prediction. Background Along with the rapid development of mobile internet and consumption financial business, a large number of cross-platform consumption financial mobile terminal applications adopt a hybrid development mode, and key transaction flows such as account inquiry, stage application, payment confirmation and the like are concentrated in the same terminal interface. In a complex network environment, the terminal is required to maintain the smoothness of page rendering and ensure the safety and real-time interactivity of the transaction process, so that higher requirements are provided for the perception of network bandwidth fluctuation and the rendering resource scheduling at the terminal side. In the prior art, the mobile terminal adopts a simple network state marking or instant speed measuring mode, for example, a unified rendering quality or a unified degradation strategy is adopted for the whole page according to a Wi-Fi/4G/5G network type or a single bandwidth measuring result, and the problems are relieved by retrying, prolonging the timeout time or integrally reducing the page refreshing frequency when a timeout is requested or a user perceives an obvious card. The technology generally does not collect the bandwidth in a continuous time-sharing manner, and also lacks modeling and prediction of the bandwidth change trend in a period of time window, and cannot perform feedforward adjustment on terminal side resources before network jitter occurs, so that uncontrollable clamping and delay are easy to occur in a critical transaction area and a non-critical area when resources compete. Therefore, the scheme aims to provide a consumption financial mobile terminal rendering scheduling method based on bandwidth prediction, and introduces two large mechanisms of local bandwidth dynamic perception and multi-region rendering thread self-adaptive allocation. On one hand, the method combines a smoothing strategy of moving average and saturation cut-off to evaluate the bandwidth fluctuation trend in real time and correct the deviation of the lower period bandwidth, and on the other hand, the method divides the screen into different functional areas and ensures that the requirement of a high priority area on the frame rate is preferentially met under the condition of limited rendering thread resources by multiple constraints such as linear mapping, segmentation correction, compression ratio, integer rounding and allowance complement. The scheme also carries out truncation and average delay calculation on the rendering response delay, and periodically adjusts the bandwidth sampling period and the number of points based on the truncation and average delay calculation, thereby forming double closed loop optimization on network fluctuation and rendering performance. Disclosure of Invention The invention provides a rendering scheduling method of a consumer financial mobile terminal based on bandwidth prediction, which facilitates solving the problems mentioned in the background art. The invention provides a rendering scheduling method of a consumer financial mobile terminal based on bandwidth prediction, which comprises the following steps: Executing time-sharing acquisition and initialization construction of network bandwidth data at a mobile terminal side to form an available downlink bandwidth sampling sequence corresponding to each scheduling period; based on the available downlink bandwidth sampling sequence, performing calculation and smoothing processing on the bandwidth change rate of adjacent sampling points to obtain a final estimation value of the local bandwidth change rate of the current scheduling period, calculating average available bandwidth according to the bandwidth sampling value in a bandwidth acquisition time window, combining the final estimation value of the local bandwidth change rate to perform prediction and lower limit correction on the lower period target bandwidth, dividing a plurality of resource areas in a terminal display page, setting a total resource amount, setting a minimum configuration value and a maximum configuration value for each resource area to form multi-area resource allocation constraint and resource demand description, constructing a mapping relation between the bandwidth and the rendering thread number of each resource area graphics processor according to the lower period target bandwidth, performing segmentation processing on the continuous resource demand number of each resource area, performing resource compression, expansion and allocation volume correction on the continuous resource demand number of each resource area acquired according to the mapping relation under the total